Zheshang Securities released a research report stating that the technological dividends from large language models are driving a new round of growth and bringing entirely new interactive experiences. The integration of AI and e-commerce is currently at a critical juncture, transitioning from "discriminative recommendation" to "generative recommendation (GR)". With the validation of Scaling Law in the recommendation field, the industry is breaking through the bottlenecks of traditional deep learning (DLRM). This not only shatters information cocoons but also reduces engineering costs through a unified end-to-end framework. E-commerce platforms are now entering a dual红利 period for enhancing user retention time and advertising click-through rates (CTR).
Generative recommendation engines utilize Large Language Models for massive product matching. Generative recommendation leverages the sequence modeling and reasoning capabilities of LLMs to significantly improve recommendation effectiveness. In February 2025, Alibaba's technical team introduced the Large User Model (LUM), which unleashes the potential for scalable recommendations while meeting the stringent requirements of industrial environments. Experiments show that LUM's performance surpasses state-of-the-art DLRMs and E2E-GR methods, demonstrating excellent scalability and reconstructing the "person-product" matching logic.
Innovation in interactive interfaces is led by Qianwen APP and Rufus, pioneering conversational shopping. Following its public beta test, Qianwen APP experienced rapid growth in Monthly Active Users (MAU). According to reports, Qianwen APP's MAU surpassed 100 million on January 14, 2026, with a new version scheduled for release on January 15, 2026. It is anticipated that Alibaba will leverage its group ecosystem to bolster the Qianwen APP entry point. The key differentiator for Qianwen compared to Doubao is expected to be Agent + Group Ecosystem synergy. Therefore, expectations are high for Qianwen's agent to integrate across the Alibaba ecosystem, including Amap (integration achieved in December 2025), Taotian, Flash Sales, Fliggy, and others. Reflecting on Amazon's AI shopping assistant Rufus, it similarly transformed the traditional search box model by allowing users to compare prices, select products, and receive suggestions through natural language queries. It provides an initial set of highly relevant SKU recommendations based on user needs, marking the formal reshaping of e-commerce traffic entry points and distribution mechanisms by generative AI.
Regarding investment targets, the report strongly recommends BABA-W (09988) and suggests monitoring companies in the industrial chain such as Focus Media (002027.SZ), SMZDM (300785.SZ), One Net One Creation (300792.SZ), Weimob (02013), Shiji Information (002153.SZ), AOK Technologies (301110.SZ), Meideng Technology (920227.BJ), Daily Interactive (300766.SZ), Focus Technology (002315.SZ), and Guangyun Technology (688365.SH).
Risk warnings include intensifying industry competition and macroeconomic policy risks.
Comments